All mice used in this study were between 7 and 10 weeks of age on a fully C57BL/6J background. Vav-Cre and Tcf7 ^fl/fl mice (from Jackson Laboratories) were maintained as previously described (33, 34). CD122-Cre mice were kindly provided by Dr. Zhongjun Dong (Tsinghua University, China). Ncr1-Cre mice were purchased from the Biocytogen Pharmaceuticals company (Beijing, China). Mice were housed in specific pathogen-free conditions under controlled temperature (22 ± 1°C) and exposed to a constant 12-hour light/dark cycle. All institutional and national guidelines for the care and use of laboratory animals were followed and all animal protocols used in this study were approved by the Institutional Animal Care and Use Committee at China Agricultural University.

Single-cell suspensions were isolated and prepared from bone marrow (BM), spleen (SP), or peripheral blood cells (PBCs) as previously described (35). For surface staining, cells were stained with fluorescence-conjugated antibodies in FACS buffer (PBS + 2% FBS). For intranuclear staining of transcription factors Tcf1, cells were fixed and permeabilized with Cytofix/Cytoperm Fixation/Permeabilization Solution Kit (554714, BD Biosciences) following the manufacturer’s instruction as previously described (36). The following fluorochrome-labeled monoclonal antibodies from eBioscience were used: CD3e (145-2C11), CD122 (TM-b1), CD49b (DX5), NK1.1 (PK136), CD11b (M1/70), CD27 (LG.7F9), Ly49A (A1), Ly49D (eBio4E5), Ly49E/F (CM4), Ly49G2 (eBio4D11), and Ly49H (3D10). Samples were acquired with FACSVerse or FACS LSRFortessa II (BD Biosciences) following the manufacturer’s instruction. The flow cytometry data were analyzed with FlowJo software (Version 10.4.0, Tree Star, Inc.). For cell sorting, cells were stained with corresponding fluorescence-conjugated antibodies and subjected to sorting on a FACS Aria II (BD Biosciences) as previously described (37).

Total RNA was extracted from sorted cells using the RNeasy Mini Kit (Qiagen) or RNeasy Micro Kit (Qiagen) followed by cDNA synthesis with FastQuant RT Kit (Tiangen) as previously described (36). Quantitative RT-PCR was carried out with SuperReal PreMix Plus SYBR Green (Tiangen) on a CFX96 Connect™ Real-Time System (Bio-Rad). The results were normalized to the expression of housekeeping gene Gapdh transcript. Differences in expression levels were calculated according to the 2^−ΔCT method. All primers used are listed in Supplementary Table 1 .

Single-cell suspensions were stained with indicated surface antibodies, and then resuspended in Annexin V binding buffer (BD Biosciences). Apoptosis assays were performed by staining cells with Annexin V and 7-aminoactinomycin D (7-AAD) (BD Biosciences) following the manufacturer’s instruction. The stained cells were acquired on a FACSVerse or FACS LSRFortessa II equipment (BD Biosciences).

For NK related genes transcriptional profiles, RNA-seq of BM and splenic NK cells subpopulations were originated from the Gene Expression Omnibus with accession no. GSE109125. The normalized counts were generated with the DESeq2 package in R (4.1.4). The clustered heatmap was performed using the “pheatmap” package in R (4.1.4) with the scale = “row” parameter.

ChIP-seq of splenic CD8⁺ T cells was originated from the Gene Expression Omnibus with accession no. GSE73239. Quality assessment of raw reads was assessed by FastQC (0.11.9) and adaptors were removed using Trim_galore (0.6.3). The processed data were mapped to the mouse reference genomes (mm10) which were downloaded from the UCSC repository using Bowtie (2.1.0). The aligned results converted to bigWig format by Bedtools (2.26.0) were visualized on the Integrated Genome Viewer from the Broad Institute. MACS (2.1.1) was used to call binding sites (peaks) relative to Input libraries with the p-value threshold set as 1 × 10⁻⁵. Finally, called peaks were annotated against the mouse reference genomes (mm10) in HOMER.

Given that there are no published Tcf1-dependent ATAC-seq in NK cells, we have initially compared ATAC-seq dataset (GSE99159) in CD4⁺CD8⁺ double positive (DP) thymocytes and several datasets (GSE119839, GSE144382, and GSE139024) in CD8⁺ cells which are associated with the Tcf1 regulation. We eventually chose the ATAC-seq data (GSE99159) of DP thymocytes, which were a high-quality data to reflect the altered chromosome accessibility in the absence of Tcf1 during naïve lymphocyte development. The quality assessment of raw reads was assessed by FastQC (0.11.9) and adaptors were removed using Trim_galore (0.6.3). The processed data were mapped to the mouse reference genomes (mm10) which were downloaded from the UCSC repository using Bowtie (2.1.0). Subsequently, the PCR duplication was eliminated by Sambamba (0.8.1). The aligned results converted to bigWig format by Bedtools (2.26.0) were visualized on the Integrated Genome Viewer from the Broad Institute. Peak calling was performed using MACS (2.1.1) with default parameter and regions of differential enrichment being tested with the DiffBind package in R (4.1.4).

We generated the NK cell-related gene set ourselves for overlapping analysis with Tcf1 regulated genes. In brief, the NK cell-related genes were retrieved genes in part from the existing gene sets associated with NK cells on the Gene Ontology Resource website (http://geneontology.org/), GSEA-MSigDB (http://www.gsea-msigdb.org/), and a linked data server (http://www.ontobee.org/), and combined a number of known-genes involved in NK cell development and/or function based on published literatures. The NK cell-related genes, Tcf1 binding gene set, and differentially accessible genes set were used for Venn analysis. Venn diagram analysis was performed using online software at the following URL: http://bioinformatics.psb.ugent.be/webtools/Venn/.

Statistical analysis was performed using Prism 7.0 (GraphPad Software) and the error bar was shown as means ± SD. The data were shown as mean ± SD in all graphs, and statistical differences were calculated using a one-tailed unpaired Student’s t-test, unless otherwise specified. The statistically significant measurements are marked as follows: *P <0.05, **P <0.01, and ***P <0.001.

To gain insights of the importance of Tcf1 (encoded by Tcf7) in the entire process of NK cell development, we went over the expression of Tcf7 from hematopoietic stem cells (HSC) to distinct NK cell subsets along with developmental phases by analyzing published data (GSE109125) from the ImmGen database. Compared with other classic NK regulators, Tcf7 exhibits dynamic expression and upregulates its expression beyond common lymphoid progenitor (CLP) stage ( Figure 1A ). Meanwhile, we found it reached a peak in CD11b^low (CD3e⁻NK1.1⁺CD27⁺CD11b⁻) iNK compartment in both bone marrow (BM) and spleen (SP) at a relatively high level in comparison with housekeeping gene expression of Gapdh or Hprt1 ( Figure 1A ). The expression of Tcf7 in distinct NK subpopulations was further confirmed by qPCR, and its expression trend was in line with high-throughput data, exhibiting high abundances in NKP (CD3e⁻CD122⁺ DX5⁻NK1.1⁻) and CD11b^low stages ( Figure 1B ). These data collectively implied the potential roles of Tcf1 in dynamically regulating early NK cell development.

Recently, Jeevan-Raj et al. have identified that Tcf1 intrinsically regulates the development of NK cells and guarantees their survival, but the precise mechanism is still unclear (31). To further elucidate stage-specific roles of Tcf1 in NK cell development, we employed three genetic models by crossing Tcf7 ^fl/fl mice with Vav-Cre (38), CD122-Cre (39) or Ncr1-Cre (40) transgenic mice to conditionally inactivate Tcf1 at HSC, NKP or iNK stage, respectively ( Figure 1C ). The deletion efficiency was further confirmed by intracellular staining (ICS) in relative subsets of NK cells ( Figures 1D, E ). Tcf1 was efficiently deleted at protein level in NKP, CD11b^low, and DP (CD3e⁻NK1.1⁺CD27⁺CD11b⁺) stages from both Tcf7 ^fl/fl Vav ^Cre/+ and Tcf7 ^fl/fl CD122 ^Cre/+ mice ( Figures 1D, E ). The comparable efficiency of Tcf1 deletion was also detected in CD11b^low, and DP subpopulations excluding NKP cells from Tcf7 ^fl/fl Ncr1 ^Cre/+ mice since the Cre recombinase is only expressed in Ncr1⁺ cells ( Figures 1D, E ). The protein level of Tcf1 in terminal mNK (CD27^low, CD3e⁻NK1.1⁺CD27⁻CD11b⁺) cells was not significantly reduced in all three conditional knockout strains owing to its negligible expression at this stage ( Figures 1D, E ).

We first examined the NK cells in BM, SP, and peripheral blood cells (PBCs) from Tcf7 ^fl/fl Vav ^Cre/+ mice. Compared with their wildtype littermate controls (Ctrls), the frequency and numbers of total NK1.1⁺ cells were substantially reduced in the absence of Tcf1 ( Figures 2A, B ). We next analyzed the early stage development of NK cell in BM with CD122 and DX5 staining. The frequency of NKP subset was notably increased, whereas the percentage of CD3e⁻CD122⁺DX5⁺NK1.1⁺ cells was remarkably decreased in BM from Tcf7 ^fl/fl Vav ^Cre/+ mice ( Figures 2C, D ). Although the absolute numbers of CD3e⁻CD122⁺DX5⁺NK1.1⁺ cells were severely diminished, there was no significant difference in the numbers of NKP and CD3e⁻CD122⁺DX5⁻NK1.1⁺ cells. Further, we found that the frequency and numbers of CD11b^low cells and transient DP NK cells were notably impaired ( Figures 2E, F ). Although the frequency of CD27^low cells was substantially increased in both BM and SP from Tcf7 ^fl/fl Vav ^Cre/+ mice, the cell numbers were remarkably reduced owing to diminished total numbers of NK1.1⁺ cells ( Figures 2E, F ). Given that Tcf1 plays a key role in the establishment of a repertoire of MHC class I-specific Ly49 receptors (28, 41), we then analyzed the expression of Ly49 receptors in splenic NK cells. The reduced expression of Ly49A and Ly49D and the increased expression of Ly49E/F and Ly49G2 were observed in splenic NK1.1⁺ cell from Tcf7 ^fl/fl Vav ^Cre/+ mice, whereas the expression of Ly49H was not altered in comparison with those from Ctrls ( Figures 2G, H ). In addition, the terminal mature receptor KLRG1 was significantly more abundant ( Figures 2G, H ). These findings collectively indicated that conditional ablation of Tcf1 at the HSC stage resulted in a blockade from NKP to iNK transition and more abundant terminal maturation of NK cell development, which were in accordance with the phenotypes in NK cells from Tcf1 germline knockout mice (31).

We next detected NK cells in the Tcf7 ^fl/fl CD122 ^Cre/+ mice. Similar to the Tcf7 ^fl/fl Vav ^Cre/+ mice, the frequency and numbers of total NK1.1⁺ cells were substantially reduced in BM, SP, and PBCs due to conditional targeting Tcf1 with CD122-Cre ( Figures 3A, B ). Accordingly, a blockade of the development from NKP to next stage was observed in BM from the Tcf7 ^fl/fl CD122 ^Cre/+ mice ( Figures 3C, D ). In both BM and SP, the percentages of CD11b^low iNK cells and DP NK cells were severely reduced and the percentage of CD27^low mNK cells was elevated, whereas the absolute numbers of each subset were impaired corresponding to diminished total numbers of NK1.1⁺ cells in the Tcf7 ^fl/fl CD122 ^Cre/+ mice ( Figures 3E, F ). The expression of Ly49 and KLRG1 receptors exhibited similar alteration in splenic NK cells from the Tcf7 ^fl/fl CD122 ^Cre/+ mice in comparison with those in the Tcf7 ^fl/fl Vav ^Cre/+ mice, except the expression of Ly49G2 was not significantly changed ( Figures 3G, H ). Together, these data suggested that Tcf1 deficiency at NKP stage caused consistent phenotypic defects in NK cells lacking Tcf1 at HSC stage.

By analyzing Tcf7 ^fl/fl Ncr1 ^Cre/+ mice, we found that the frequency and numbers of total NK1.1⁺ cells were significantly reduced in BM, SP, and PBCs in comparison to those from their littermate Ctrls ( Figures 4A, B ). The decreased frequency and numbers in BM NK cells from the Tcf7 ^fl/fl Ncr1 ^Cre/+ mice were comparable with those in the Tcf7 ^fl/fl Vav ^Cre/+or Tcf7 ^fl/fl CD122 ^Cre/+ mice, but the phenotypic defects on impaired frequency and numbers of NK cell in SP or PBCs were modest ( Figures 4A, B ). Similar to the Tcf7 ^fl/fl Vav ^Cre/+and Tcf7 ^fl/fl CD122 ^Cre/+ mice, the frequency of NKP subset was notably increased and the percentage of CD3e⁻CD122⁺DX5⁺NK1.1⁺ cells was decreased in BM from the Tcf7 ^fl/fl Ncr1 ^Cre/+ mice ( Figures 4C, D ). The absolute numbers of NKP cells were not altered, whereas the CD3e⁻CD122⁺DX5⁺ NK1.1⁺ cells were diminished ( Figures 4C, D ). The similar defective phenotypes in frequency and numbers of CD11b^low, DP, and CD27^low NK cells were exhibited in comparison with those from the Tcf7 ^fl/fl Vav ^Cre/+and Tcf7 ^fl/fl CD122 ^Cre/+ mice ( Figures 4E, F ). For the expression of Ly49 receptors, we found that conditional ablation of Tcf1 in only Ncr1⁺ cells led to decreased expression of Ly49A, Ly49E/F, and Ly49G2 and increased expression of KLRG1, but no alteration in Ly49D. These results are in line with the findings from Dr. Barbara Kee’s group by using their Tcf7 ^fl/fl Ncr1 ^Cre/+ strain (32), but not consistent with those from mice lacking Tcf1 at NKP stage, particularly in expression of Ly49D, Ly49E/F, and Ly49G2 ( Figures 4G, H ). Collectively, compared with the Tcf7 ^fl/fl Vav ^Cre/+ and Tcf7 ^fl/fl CD122 ^Cre/+ mice, the Tcf7 ^fl/fl Ncr1 ^Cre/+ mice exhibited similar phenotypes in BM NK cells but showed modest defects and distinct expression patterns of Ly49 receptors in splenic NK cells.

Given impaired survival was found in NK cells from Tcf1 germline knockout mice due to excessive expression of GZMB (31), we next extended the apoptosis assay in distinct subsets of BM and SP NK cells from all three strains. We found the percentages of Annexin V⁺ cells were significantly increased in all NK subpopulations from the Tcf7 ^fl/fl Vav ^Cre/+, Tcf7 ^fl/fl CD122 ^Cre/+, and Tcf7 ^fl/fl Ncr1 ^Cre/+ mice ( Figures 5A – F ), indicating Tcf1 was crucial for survival of NK cells at each developmental phase.

To further interrogate how Tcf1 modulates the key regulators in early NK cell development, we constructed a Tcf1-dependent transcriptional network by combining analysis of existing ATAC-seq and ChIP-seq data with NK regulatory genes. First, we established our “NK cell-related gene set” by combining the NK cell-associated genes collected from existing NK gene sets from three websites with established NK regulatory genes retrieved collected from published literatures (described in details in Materials and Methods). A Venn diagram analysis revealed that 110 of NK-cell related genes were directly bound by Tcf1 and their chromatin accessibility relied on Tcf1 ( Figure 6A ). In addition, 26 of NK-cell related genes which showed irrelevance with chromatin accessibility fell into the Tcf1 binding gene set, whereas 60 of NK cell-related genes exhibited differential chromatin accessibility in absence of Tcf1, but not directly bound by Tcf1 ( Figure 6A ). Based on previous knowledge, we further inspected a number of essential NK regulatory genes loci and analyzed their Tcf1 binding peaks and ATAC-seq peaks’ distribution, including Eomes, Ets1, Gata3, Gzmb, Id2, Ikzf1, Ikzf2, Nfil3, RunX3, Sh2d1a, Slamf6, Tbx21, Tox, Zeb2, and Spi1. At the promoter and regulatory regions of Tcf1 binding genes Eomes, Ets1, Gata3, Id2, Ikzf1, Nfil3, RunX3, Sh2d1a, and Tox, the alteration of chromatin accessibility was tightly associated with the absence of Tcf1 ( Figure 6B ). Meanwhile, Gzmb, Slamf6, and Tbx21 were directly bound by Tcf1, whereas the chromatin accessibility did not show significant difference ( Figure 6B ). By contrast, we observed Ikzf2 and Zeb2 were indirectly regulated by Tcf1 due to the altered chromatin accessibility caused by Tcf1 deficiency ( Figure 6B ). Given that Spi1 has a critical function in early NK cell development, we exhibited its IGV for comparison even though neither was it bound by Tcf1 nor its chromatin accessibility was altered ( Figure 6B ). These results collectively suggested that Tcf1 is in charge of transcription programs corresponding to early NK cell development in a direct or indirect manner.

We next confirmed the expression of genes referred in Figure 6B in NKP, CD11b^low, DP, and CD27^low cells from the Tcf7 ^fl/fl CD122 ^Cre/+ mice by qPCR. The results indicated most of them were downregulated in all developmental stages, namely, Eomes, Ets1, Gata3, Ikzf1, Ikzf2, Runx3, Sh2d1a, and Slamf6, reflecting Tcf1 positively regulated their expression in NK cells ( Figure 7A ). Contrastingly, Spi1 and Gzmb were upregulated at all developmental stages of NK cell in absence of Tcf1, implying Tcf1 negatively regulated their expression in NK cells ( Figure 7A ). Nfil3 exhibited a slight elevation in NKP stage and its expression was slightly impaired in following developmental phases due to ablation of Tcf1 ( Figure 7A ). Tox was downregulated only in NKP cells caused by Tcf1 deficiency, but no significant alteration was observed in CD11b^low, DP, and CD27^low cells ( Figure 7A ). The expression of Tbx21 was significantly lower in Tcf1-deficient NKP and CD27^low cells than that from Ctrls, but only slightly downregulated in CD11b^low and DP subsets ( Figure 7A ). Despite the differential gene expression in NKP might be caused by the loss of NK progenitors in Tcf1-deficient mice, the expression of many genes exhibits the same trends in following more mature NK cells. It is worthy mentioning that the altered expression of Tox and Nfil3 was only detected in NKP cells, but no significant changes in the following stages, suggesting that we have to cautiously consider their aberrant expression in NKP stage due to Tcf1 deficiency. Based on our bioinformatic analysis above and the gene expression results generated from the Tcf7 ^fl/fl CD122 ^Cre/+ mice, we constructed a working model to summarize the Tcf1 related regulatory programs linking to early NK cell development ( Figure 7B ). The dynamic expression of Tcf1 starts at CLP stage and reaches a peak in NKP and CD11b^low NK cells. During NK cell maturation, Tcf1 turns down its expression in DP subset and remains negligible level in CD27^low terminal mNK cells. All these four NK subpopulations exhibit the defective development when Tcf1 is deactivated prior to NKP phase. A number of essential NK governing genes are dysregulated in Tcf1-deficient NK cells, reflecting Tcf1 serves as their modulator in a direct or indirect manner. Thus, we identified the essential targets of Tcf1 and uncovered a Tcf1 associated regulatory network in maintaining early NK cell development.